Fire monitoring and early warning device for new energy power generation net cage and combiner box

Abstract

The invention discloses a fire monitoring and early warning device for a new energy power generation net cage and a combiner box. The device comprises a sensing network module, an edge calculation and data fusion gateway for intelligent research and judgment, and a hierarchical early warning and linkage control module for executing hierarchical response, wherein the sensing network module is integrated with a plurality of heterogeneous sensors such as an air-breathing extremely-early smoke detection sensor, a distributed optical fiber temperature measurement sensor, a wide-spectrum arc light detection sensor and a characteristic gas monitoring sensor, so that synchronous acquisition of multi-dimensional early symptoms of a fire disaster is realized; according to the invention, real-time fusion analysis is carried out on multi-source information through the edge computing gateway, a continuously quantified fire risk index is generated, and the early warning control module is driven to automatically execute linkage response from local alarm, active ventilation load reduction to emergency power failure and fixed-point suppression according to the risk level. Therefore, the problems that in the prior art, the complex electrical fire recognition capacity is insufficient, the false alarm and missing alarm rate is high, and comprehensive risk judgment and active intervention are lacked are solved.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for new energy power generation equipment, and in particular to a fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes. Background Technology

[0002] With the rapid development of the new energy power generation industry, integrated power generation grid boxes (containing converters, controllers, etc.) and combiner boxes are used in photovoltaic power stations, wind farms and other scenarios. These devices have dense internal electrical components, large operating current, and are exposed to harsh outdoor environments for a long time. They pose a significant risk of fire due to insulation aging, poor contact, overload and electric arc. At present, fire warning for such devices is mainly achieved by installing a single or a few traditional sensors, such as point smoke detectors, heat detectors or arc fault circuit breakers. These technical solutions usually operate independently, have fixed alarm thresholds, and can only respond to a single feature of fire development (such as smoke concentration, temperature or a specific electric arc).

[0003] The internal environment of new energy power generation equipment is complex, and the development from electrical faults to fires is a multi-stage and multi-symptom process. Individual sensors are susceptible to factors such as normal equipment start-up and shutdown, changes in ambient temperature and humidity, and dust interference, leading to a high false alarm rate. Furthermore, for rapidly developing faults (such as high-current arcing) or hidden early hazards (such as localized overheating or slow pyrolysis of insulation materials), the response of a single sensor may be delayed, resulting in missed alarms or insufficient warning time, failing to provide sufficient safety time for operator intervention or the activation of automatic response systems. Moreover, independent alarms from individual sensors lack comprehensive analysis of multi-source information, making it difficult to accurately determine the authenticity, severity, and development trend of faults, thus failing to meet high safety requirements for the accuracy and reliability of early warnings.

[0004] Therefore, in response to the problems mentioned above, this invention proposes a fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes. Summary of the Invention

[0005] To overcome the shortcomings of existing fire monitoring technologies, such as insufficient early and multi-symptom identification capabilities for complex electrical fires, high false alarm and missed alarm rates, and lack of comprehensive risk level assessment, this invention proposes a fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes, which can realize early warning of multiple symptoms and intelligent handling.

[0006] The technical solution of this invention is: a fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes, comprising:

[0007] The sensor network module is installed inside the electrical compartment of the monitored power grid box, inside the combiner box, and at its key cable connections. The distributed sensor network module includes multiple sensors for synchronously collecting multidimensional fire characteristic signals.

[0008] The edge computing and data fusion gateway communicates with the distributed sensor network module and has a built-in edge computing unit and multi-source information fusion algorithm model.

[0009] The hierarchical early warning and linkage control module is connected to the edge computing and data fusion gateway to output hierarchical early warning signals and execute preset linkage commands based on the risk level after fusion.

[0010] The sensor network module includes at least the following:

[0011] The aspirating early smoke detection unit has a sampling pipeline network that extends to cover the top space of the power grid box and electrical cabinet and the busbar area of ​​the combiner box, and is used to detect the concentration of submicron smoke particles and their changing trends.

[0012] The fiber optic temperature measurement unit has its temperature-measuring optical cable laid close to the surface of the power module heat dissipation substrate, the DC busbar, and the incoming and outgoing cables of the combiner box inside the power grid box, enabling continuous spatial temperature field monitoring. The fiber optic temperature measurement unit adopts a distributed temperature measurement system based on the principles of Raman scattering or Brillouin scattering, with a spatial resolution of 0.5 meters and a temperature measurement accuracy of ±1℃. It is equipped with two or more temperature alarm thresholds: the first threshold is the absolute temperature value, used to monitor overheating; the second threshold is the temperature rise rate per unit length, used to monitor local rapid temperature rise.

[0013] The broadband arc light detection unit has an optical sensor pointing to the circuit breaker, contactor, and junction box terminal block of the power grid box to monitor transient arc light signals in the ultraviolet to visible light band. The broadband arc light detection unit includes an optical sensor and a current change detection circuit. The optical sensor is equipped with a filter of a specific wavelength to suppress ambient light interference. The edge computing and data fusion gateway only confirms a valid arc event when the arc light signal received from the optical sensor exceeds a set threshold and the current change detection circuit detects a sudden change in the current of the corresponding circuit.

[0014] The characteristic gas monitoring unit has a gas sampling probe installed in the upper part of the equipment compartment to detect the concentration of characteristic gases generated by the pyrolysis of insulating materials. The characteristic gas monitoring unit uses an electrochemical or photoionization sensor, and the types of gases it monitors include at least carbon monoxide and the total amount of volatile organic compounds. It is also equipped with a composite alarm strategy based on gas concentration and concentration change rate.

[0015] The workflow of the edge computing and data fusion gateway is as follows:

[0016] A1 receives and synchronizes the raw data sent by each sensor in time;

[0017] A2 extracts features from the original data to obtain multi-dimensional feature vectors including but not limited to the rate of increase in smoke concentration, local temperature gradient, frequency of arc events and the ratio of energy to the concentration of characteristic gases.

[0018] A3 inputs multi-dimensional feature vectors into a trained multi-source information fusion algorithm model, which is based on weighted evidence theory or deep neural networks, and outputs a comprehensive fire risk index and corresponding main risk factor identifiers.

[0019] A4, based on the threshold range of the fire risk index, sends the corresponding risk level signal to the hierarchical early warning and linkage control module.

[0020] Preferably, the multi-source information fusion algorithm model adopts an adaptive weight allocation mechanism, in which the weight coefficient of at least one sensor is adjusted according to the equipment operating status and environmental background noise. When the equipment is running under high load, the weight of the fiber optic temperature measurement unit and the broadband arc light detection unit is increased; during the restart phase after equipment shutdown or maintenance, the weight of the aspirating early smoke detection unit and the characteristic gas monitoring unit is increased.

[0021] Preferably, the edge computing and data fusion gateway includes a device thermal model, which uses ambient temperature and device load current as inputs to predict the normal temperature rise curve of the device; wherein the edge computing and data fusion gateway compares the actual temperature measured by the fiber optic temperature measurement unit with the temperature predicted by the thermal model in real time, and when the deviation continues to exceed the set range and measurement error is eliminated, a potential fault warning is triggered even if the absolute temperature does not exceed the limit.

[0022] Preferably, the hierarchical early warning and linkage control module includes at least three levels of early warning response:

[0023] Primary warning (risk index 30-60): Local audible and visual alarms are activated, and the warning information is uploaded to the monitoring backend.

[0024] Intermediate warning (risk index 61-85): Strengthen local alarm, automatically activate directional ventilation devices in the equipment compartment for heat dissipation and dilution, and can choose to reduce the output power of the power generation unit via remote command;

[0025] Advanced warning (risk index 86-100) triggers the highest level audible and visual alarm, sends an emergency shutdown request command to the monitoring backend, and activates the compressed air foam or fine water mist device pre-installed above the critical equipment. It also controls the solenoid valve of the compressed air foam or fine water mist device pre-installed above the critical equipment (such as converter modules and busbars) to activate the targeted fire suppression. The activation signal of the solenoid valve is hard-wired and interlocked with the trip signal of the upstream circuit breaker of the generator box or combiner box to ensure that the relevant electrical circuit has been reliably disconnected before or at the same time as the fire suppression is implemented.

[0026] Preferably, the device also includes a data communication and remote monitoring module, which uploads multi-dimensional feature data, fire risk index, alarm events and equipment status information processed by the edge computing and data fusion gateway to the power plant control center via industrial Ethernet or wireless private network, and can receive parameter setting, model update and diagnostic test instructions issued by the control center.

[0027] The beneficial effects of this invention are:

[0028] 1. This invention constructs a heterogeneous sensor network that integrates aspirating smoke detection, distributed fiber optic temperature measurement, broadband arc light detection, and characteristic gas monitoring. This enables simultaneous and multi-dimensional sensing of four key physicochemical signals generated during the incubation of electrical fires: smoke particles, spatial temperature field distribution, transient arc light, and characteristic gases. It captures early and multi-type fire signs that cannot be covered by traditional single sensors from different physical levels, solving the problem of insufficient early and multi-sign recognition capabilities of existing technologies for complex electrical fires.

[0029] 2. This invention introduces an edge computing gateway with a multi-source information fusion algorithm model. This model can perform time synchronization, feature extraction, and adaptive weight fusion calculation on heterogeneous data from different sensors, transforming discrete and potentially contradictory signals into a continuous, quantitative, comprehensive fire risk index. This enables intelligent and highly reliable comprehensive judgment of the authenticity of faults, risk levels, and development trends, thereby significantly reducing the false alarm and missed alarm rates caused by environmental interference or misjudgment by a single sensor. Attached Figure Description

[0030] Figure 1 The diagram shown is a schematic representation of the system framework of this invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figure 1 This invention provides an embodiment of a fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes, comprising:

[0033] In this embodiment, the sensor network module will be described in detail:

[0034] (1) The aspirating early smoke detection unit is installed on the top of the electrical compartment (heat and smoke accumulation area) and inside the combiner box of a single power generation grid box. A sampling capillary network with a diameter of about 8-10 mm is laid. The network adopts a "T" or "U" shaped layout. The spacing of the sampling holes is optimized according to the volume of the protected area. A high-sensitivity laser particle counting detector is connected to the end of the sampling tube. The detector has a built-in air pump to continuously draw air samples from the protected area into the analysis chamber.

[0035] This unit is highly sensitive to submicron (0.003-1.0μm) combustion and gasification particles, and its alarm threshold (typically set at 0.08% obs / m attenuation) is far lower than that of traditional point-type smoke detectors (approximately 3-5% obs / m). Even before visible smoke is produced, as the insulating material begins slow pyrolysis due to overheating, this unit can detect a sustained abnormal rise in particle concentration, providing the earliest warning signal. By analyzing the rate of increase in concentration at the gateway, it can further distinguish between transient dust interference and a continuous pyrolysis process.

[0036] (2) The fiber optic temperature measurement unit adopts a distributed temperature measurement system based on the Raman scattering principle. The temperature measurement optical cable is selected to be high temperature resistant, compression resistant, and corrosion resistant. The laying path follows the principle of "close to the heat source". High temperature resistant cable ties or thermal conductive adhesive are used to tightly bind the optical cable to the heat sink surface of the power device in the generator box, the surface of the DC support capacitor and the insulation sheath of the main DC bus copper bus, as well as the fuses, circuit breaker output terminals and main bus in each branch box. The optical cable path forms a continuous loop, with both ends connected to the temperature measurement host. In this way, true "line" and "surface" temperature monitoring is achieved, with no measurement blind spots. This unit can not only display the absolute temperature of each measurement point (the typical spatial resolution is 0.5m or 1m), but also draw the temperature gradient distribution map of the entire monitoring path in real time. Such hot spots can be detected when the overall ambient temperature is not high. The system also employs dual criteria, including temperature threshold alarms (e.g., setting the busbar alarm threshold to 85℃ and the power device heatsink alarm to 90℃) and temperature rise rate alarms (e.g., setting an alarm if the temperature rise rate at any point exceeds 10℃ / minute). The latter is highly sensitive to rapid thermal runaway caused by capacitor breakdown, internal short circuits, etc.

[0037] (3) The broadband arc light detection unit includes an optical sensor and a matching current transformer. The optical sensor is a photodiode or photomultiplier tube, with a specific filter combination installed in front (such as focusing on monitoring the ultraviolet band 320-380nm and specific visible light bands) to suppress interference from continuous background light such as sunlight and lighting. The sensor probes are installed in the circuit breaker compartment of the power grid box, above the contactor, and directly in front of the terminal block of the combiner box to ensure that the field of view covers the key connection points that may generate arcs. A high-precision current transformer is installed on the main circuit (or branch circuit) corresponding to each probe. This unit achieves high-reliability identification of arc faults by using the AND logic judgment of light intensity change and current change. When an arc occurs, it will generate strong light radiation and sudden current change in a very short time (microseconds to milliseconds). The edge gateway only confirms a valid arc event when the sensor detects a sudden increase in light intensity exceeding 100 times the background value within 1 ms (the threshold is adjustable) and the CT detects a change rate of the same circuit current exceeding 50% within 2 ms (such as a sudden change from 100A to 150A or a sudden drop to 50A). This dual criterion can effectively eliminate external light interference such as welding and camera flashes, as well as arc light during normal opening and closing of switches, thereby significantly reducing the false alarm rate of arc events.

[0038] (4) The characteristic gas monitoring unit is installed on the top of the power generation grid box and the combiner box (where gas tends to accumulate). The detector has two independent sensing elements built in, including an electrochemical carbon monoxide sensor and a photoionization volatile organic compound (TVOC) sensor. The sampling method can be natural diffusion. For boxes with good sealing, a micro air pump can be added for active sampling. When the insulating material (such as epoxy resin, polyester film, cable insulation layer) is overheated but not ignited by open flame, it will release specific gaseous products through thermal decomposition. CO and various hydrocarbon VOCs are typical examples. Monitoring the concentration baseline of these characteristic gases and their changing trends is the chemical basis for judging whether the insulating material has suffered continuous thermal damage. This complements the smoke detection. Some materials produce gas first and then smoke after pyrolysis, while others do the opposite. The system sets composite alarm conditions, such as: CO concentration > 25 ppm or TVOC concentration > 15 ppm, while also satisfying the concentration change rate > 3 ppm / min. This helps to distinguish between trace gases that may exist during normal operation of the equipment and the continuously deteriorating pyrolysis process.

[0039] In this embodiment, the edge computing and data fusion gateway will be described in detail:

[0040] The gateway runs a high-precision Network Time Protocol (NTP) or Precision Time Protocol (PTP) to assign a uniform time stamp (accuracy <10ms) to all asynchronously incoming sensor data, and then performs filtering (such as Kalman filtering), noise reduction and standardization on the raw data.

[0041] Then, multidimensional feature vectors are calculated by sliding them across a fixed time window (e.g., 10 seconds, 1 minute) from the preprocessed data. For example:

[0042] Extract the current smoke concentration value and the rate of increase in concentration over the past minute from the ASD data.

[0043] Extract the highest temperature point value, the temperature rise rate over the past 30 seconds from the DTS data, and the area percentage of the high-temperature region (>70℃).

[0044] Extract the count of valid arc events in the past 5 minutes and the peak light intensity of the most recent arc from the arc data.

[0045] Extract the current CO concentration, current TVOC concentration, and CO change rate over the past 2 minutes from the gas data.

[0046] This gateway integrates a multi-source information fusion algorithm model, and two examples are provided here for illustration:

[0047] (1) A fusion model based on the improved DS evidence theory is adopted.

[0048] Each sensor's feature subset is considered as an "evidence body." For example, {current smoke concentration value, concentration rise rate over the past minute} is considered "smoke evidence," {temperature rise rate over the past 30 seconds at the highest temperature point, area percentage of high-temperature regions (>70℃)} is considered "temperature evidence," {count of effective arc events over the past 5 minutes, peak light intensity of the most recent arc} is considered "arc evidence," and {CO change rate over the past 2 minutes, current TVOC concentration} is considered "gas evidence." An identification framework is defined for each evidence body: {F (fire), P (potential fault), N (normal)}. Feature values ​​are transformed into basic probability assignments for each proposition using membership functions or historical statistical methods. A key improvement is the introduction of adaptive weights: the weighting factors dynamically depend on the equipment operating state (S) and environmental background (E), such as weighting factor (temperature) = f (load current, ambient temperature), weighting factor (smoke) = g (operating time after equipment cleaning, ambient dust concentration). Then, the Dempster combination rule is used to fuse the weighted evidence, and finally the confidence intervals for "F (fire)", "P (potential fault)" and "N (normal)" are calculated. The comprehensive fire risk index can be defined as: Comprehensive fire risk index = Bel(F)×100 + [Bel(F)+Pl(F)] / 2×50, where Bel is the confidence function and Pl is the likelihood function. This formula makes the results more inclined to conservative alarms.

[0049] (2) A fusion model based on deep learning (CNN-LSTM) is adopted.

[0050] The model inputs a multidimensional time series matrix by organizing multidimensional feature vectors into a time series (e.g., the past 10 time steps). The front end of the model uses a one-dimensional convolutional neural network (CNN) layer to extract local spatial correlations between features from different sensors (e.g., the temporal proximity of temperature rise and electric arc). The output of the CNN is then fed into a long short-term memory (LSTM) network layer to learn the temporal dependence and sequential patterns of symptom appearance during fire evolution. The output layer is a fully connected layer using the Softmax activation function, directly outputting the probability of belonging to one of four categories: {"normal", "low risk", "medium risk", "high risk"}. The comprehensive fire risk index can be mapped to a weighted sum of category probabilities, for example: Comprehensive fire risk index = P(low risk) × 30 + P(medium risk) × 65 + P(high risk) × 100. This model requires training with a large amount of historical fault data (or high-fidelity simulation data).

[0051] The gateway also includes a device thermal model, which serves as an auxiliary diagnostic tool. The gateway has a built-in simplified thermodynamic model. For key monitored power modules or busbars, the model calculates their theoretical temperature online based on real-time measured ambient temperature and load current, combined with the thermal resistance and thermal capacity parameters of that part (fitted from factory test or initial operating data). It continuously compares ΔT = actual temperature - theoretical temperature. If ΔT remains positive and exceeds a threshold (e.g., +15℃), it indicates the presence of an additional abnormal heat source (e.g., increased contact resistance). Even if the actual temperature has not yet reached the alarm threshold, the system will generate a potential thermal fault warning event, prompting maintenance personnel to inspect the device.

[0052] In this embodiment, the hierarchical early warning and linkage control module is described in detail:

[0053] This module is divided into a three-level early warning response mechanism, specifically including:

[0054] A preliminary warning (comprehensive fire risk index of 30-60) indicates the presence of early-stage hazards or minor anomalies. Upon warning, the on-site alarm light flashes yellow slowly (e.g., 1Hz), the alarm emits an intermittent low-frequency buzzer, and simultaneously, a detailed warning report is sent via the communication module to the power plant's centralized monitoring system (SCADA) or cloud platform. This report includes the risk index, key risk factors (such as "high rate of temperature rise" or "slow rise in CO concentration"), relevant sensor readings, and equipment locations.

[0055] A medium-level warning (comprehensive fire risk index of 61-85) indicates that the risk is developing and there is a high probability of it evolving into a fire. After the warning, the on-site alarm escalates to a red flashing light (e.g., 2Hz) and a continuous high-frequency buzzer. Ventilation fans installed on the top or side walls of the equipment compartment automatically start to enhance air circulation, aid heat dissipation, and dilute any potentially accumulated flammable gases. Simultaneously, the gateway sends a "recommended load reduction" command to the main controller of the power generation unit, such as actively limiting the current output power to 60%-80% of the rated value, thereby reducing the electrical load, mitigating the risk of overheating or overcurrent at the source, and buying time for manual intervention.

[0056] A high-level early warning (comprehensive fire risk index of 86-100) indicates a fire has occurred. Upon warning, a continuous red, bright, and piercing alarm is triggered. The gateway immediately sends the highest-level "Emergency Fire Shutdown" request and simultaneously issues a trip command to the upstream circuit breaker of the generator set or combiner box, forcibly cutting off the power supply. Simultaneously with the power-off command (or after a brief confirmation delay of <500ms), a targeted fire suppression operation is initiated. The solenoid valves of the compressed air foam (CAFS) or water mist fire suppression devices in the corresponding protected area (such as a combiner box or generator compartment) are opened. The extinguishing agent is precisely sprayed through pre-laid pipes and nozzles (directly above critical equipment) to rapidly suppress the initial fire. Simultaneously, the activation circuit of the fire suppression solenoid valve is connected in series with the normally closed auxiliary contact of the circuit breaker (or an independent tripping status sensor) to ensure that the fire suppression device is only activated after confirming that the circuit breaker has tripped and the power supply has been cut off, thus absolutely preventing fire suppression while the circuit is energized.

[0057] In this embodiment, the data communication and remote monitoring module will be described in detail:

[0058] This module uploads real-time status, alarm events, and risk trend data to the fire safety monitoring cloud platform located in the central control center. It also provides a visual human-machine interface, displaying real-time risk heat maps, historical data curves, and alarm statistics for all monitoring equipment across the site. Maintenance personnel can use the platform to configure parameters, upgrade software, calibrate sensor zero points, and perform remote diagnostic tests on any remote edge gateway, significantly reducing on-site maintenance costs.

[0059] This invention provides a comparative example:

[0060] This example designs a simulation experiment, simulating a 500kW rated power photovoltaic power generation grid box (containing converter modules, DC busbars, etc.), and compares it with two existing technical solutions. Specifically:

[0061] Comparative Example 1 is a traditional solution, which only installs a point-type heat detector (alarm threshold 85℃) and a point-type smoke detector. Comparative Example 2 is an improved traditional solution, which installs point-type heat and smoke detectors and adds an independent arc fault detection device.

[0062] This example sets up three fault scenarios. Scenario 1 involves connecting a variable resistor at a bolt connection on the DC busbar to simulate a slow increase in contact resistance due to loosening. Scenario 2 places a small heating element on the back of the insulation board near the power module to simulate the continuous pyrolysis of the insulation material caused by overheating of nearby components. Scenario 3 uses an adjustable-gap discharge device at the terminal block to simulate intermittent series arcs caused by vibration or contamination. These scenarios are used to verify the effectiveness of each scenario.

[0063] Experimental scenario Injection Fault Description Comparative Example 1 Comparative Example 2 This invention Scene 1 After the injection failure, the temperature at the connection point rose from 40°C at a rate of approximately 3°C / minute. Warning time > 60 minutes (until temperature approaches 85℃); only "high temperature" is reported, location cannot be determined. Warning time > 60 minutes; only "high temperature" is reported, location cannot be determined. The warning time is 12 minutes; the risk identification is "abnormal temperature rise rate of busbar at location XX", accurately locating the risk. Scene 2 The heating element raises the temperature of the insulating board to ~150℃, generating pyrolysis gases and trace amounts of smoke. The warning period is approximately 45 minutes; the risk identification report indicates "smoke," the cause of which is unknown. The warning period is approximately 45 minutes; the risk identification report indicates "smoke," the cause of which is unknown. The warning time is 8 minutes; the risk identification is "abnormally high concentration of pyrolysis gas from insulating materials". Scene 3 It generates intermittent low-energy electric arcs lasting approximately 50 ms with intervals of 2-5 seconds. No alarm was triggered (no open flame, high temperature, or large amount of smoke). The warning time is 5 minutes; the risk identification report is "arc fault", but it is susceptible to interference and false alarms. The warning time is 3 minutes; the risk identification is "Intermittent electric arc confirmed in XX circuit". Long-term reliability testing 168 hours of normal and disruptive operation There were 3 false alarms (2 due to dust disturbance, and 1 due to the high temperature in the computer room during the summer). There were 5 false alarms (including 2 false triggers of the arc detection due to sunlight flicker). 0 false alarms

[0064] As shown in the table above, by simulating typical electrical fire hazard scenarios such as slow overheating due to poor contact, pyrolysis of insulating materials, and intermittent electric arcs, the actual performance of the traditional temperature and smoke detection scheme, the improved scheme with added electric arc detection, and the multi-sensor fusion intelligent scheme of the present invention were compared. The results show that the present invention can achieve the earliest warning (as early as tens of minutes in advance) in all single fault scenarios, accurately identify the type and location of the hazard factor, and maintain zero false alarms in an anti-interference test of up to 168 hours, which is significantly better than the multiple false alarms of the comparison scheme.

[0065] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes, characterized in that, Including: The sensor network module is installed inside the electrical compartment of the monitored power grid box, inside the combiner box, and at its key cable connections. The distributed sensor network module includes multiple sensors for synchronously collecting multidimensional fire characteristic signals. The edge computing and data fusion gateway communicates with the distributed sensor network module and has a built-in edge computing unit and multi-source information fusion algorithm model. The hierarchical early warning and linkage control module is connected to the edge computing and data fusion gateway to output hierarchical early warning signals and execute preset linkage commands based on the risk level after fusion. The sensor network module includes at least the following: The aspirating early smoke detection unit has a sampling pipeline network that extends to cover the top space of the power grid box and electrical cabinet and the busbar area of ​​the combiner box, and is used to detect the concentration of submicron smoke particles and their changing trends. The fiber optic temperature measurement unit has its temperature measurement optical cable closely attached to the surface of the heat dissipation substrate of the power module inside the power grid box, as well as the DC busbar and the incoming and outgoing cables of the combiner box, to achieve continuous spatial temperature field monitoring. The broadband arc light detection unit has its optical sensors pointing towards the circuit breakers, contactors, and junction box terminal blocks of the power grid box to monitor transient arc light signals in the ultraviolet to visible light band. The characteristic gas monitoring unit has a gas sampling probe installed in the upper part of the equipment compartment to detect the concentration of characteristic gases generated by the pyrolysis of insulating materials; The workflow of the edge computing and data fusion gateway is as follows: A1 receives and synchronizes the raw data sent by each sensor in time; A2 extracts features from the original data to obtain multi-dimensional feature vectors including but not limited to the rate of increase in smoke concentration, local temperature gradient, frequency of arc events and the ratio of energy to the concentration of characteristic gases. A3 inputs multi-dimensional feature vectors into a trained multi-source information fusion algorithm model, which is based on weighted evidence theory or deep neural networks, and outputs a comprehensive fire risk index and corresponding main risk factor identifiers. A4, based on the threshold range of the fire risk index, sends the corresponding risk level signal to the hierarchical early warning and linkage control module.

2. The fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 1, characterized in that: The fiber optic temperature measurement unit adopts a distributed temperature measurement system based on the principles of Raman scattering or Brillouin scattering. Its spatial resolution is 0.5 meters, the temperature measurement accuracy is ±1℃, and it is equipped with two or more temperature alarm thresholds. The first threshold is the absolute temperature value, which is used to monitor overheating; the second threshold is the temperature rise rate per unit length, which is used to monitor local rapid temperature rise.

3. The fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 2, characterized in that: The broadband arc light detection unit includes an optical sensor and a current surge detection circuit, wherein the optical sensor is equipped with a filter of a specific wavelength to suppress ambient light interference.

4. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 3, characterized in that: The edge computing and data fusion gateway only confirms a valid arc event when it simultaneously receives arc signals from optical sensors exceeding a set threshold and the current surge detection circuit detects a sudden change in the current of the corresponding circuit.

5. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 4, characterized in that: The characteristic gas monitoring unit employs an electrochemical or photoionization sensor, and the types of gases it monitors include at least carbon monoxide and the total amount of volatile organic compounds. It is also equipped with a composite alarm strategy based on gas concentration and concentration change rate.

6. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 5, characterized in that: The multi-source information fusion algorithm model adopts an adaptive weight allocation mechanism, in which the weight coefficient of at least one sensor is adjusted according to the equipment operating status and environmental background noise. When the equipment is running under high load, the weight of the fiber optic temperature measurement unit and the broadband arc light detection unit is increased; during the restart phase after equipment shutdown or maintenance, the weight of the aspirating early smoke detection unit and the characteristic gas monitoring unit is increased.

7. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 6, characterized in that: The edge computing and data fusion gateway includes a device thermal model, which uses ambient temperature and device load current as inputs to predict the normal temperature rise curve of the device. The edge computing and data fusion gateway compares the actual temperature measured by the fiber optic temperature measurement unit with the temperature predicted by the thermal model in real time. When the deviation continues to exceed the set range and measurement errors are eliminated, a potential fault warning is triggered even if the absolute temperature does not exceed the limit.

8. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 7, characterized in that, The hierarchical early warning and linkage control module includes at least three levels of early warning response: Primary warning: The local audible and visual alarm is activated, and the warning information is uploaded to the monitoring backend. Intermediate warning, enhanced local alarm, automatic activation of directional ventilation devices in the equipment compartment for heat dissipation and dilution, and the option to reduce the output power of the power generation grid box via remote command; Advanced warning triggers the highest level audible and visual alarm, sends an emergency shutdown request command to the monitoring backend, and activates the compressed air foam or fine water mist device pre-installed above the critical equipment.

9. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 8, characterized in that: The nozzles of the compressed air foam or fine water mist device are controlled by a solenoid valve. The control logic of the solenoid valve is managed by a hierarchical early warning and linkage control module, and its start signal is interlocked with the circuit breaker trip signal of the power grid box or combiner box to ensure that the relevant electrical circuits have been cut off before or at the same time as the fire suppression is implemented.

10. A fire monitoring and early warning device for new energy power generation grid boxes and combiner boxes according to claim 9, characterized in that: The device also includes a data communication and remote monitoring module, which uploads multi-dimensional feature data, fire risk index, alarm events and equipment status information processed by edge computing and data fusion gateway to the power plant control center via industrial Ethernet or wireless private network, and can receive parameter setting, model update and diagnostic test instructions issued by the control center.

Images